realistic behavioral model for hierarchical coordinated ... · abstract—in real-time strategy...

Abstract—In real-time strategy games, a large number of

units are continuously trying to reach their strategic goal while

following a long chain of commands, and trying to keep on with

strict multi-level hierarchies and adapting to changing

formation topology orders in accordance to situation state and

local environmental features. In this paper we present a real-

time model, capable of simulating the propagation of orders

through a long chain of command, starting at a brigade level,

down through 5 levels to individual units. Each unit, while

trying to satisfy the overall brigade goals, autonomously must

design its local optimum path, avoiding obstacles and

impassable terrain, through dynamic path refinement and

realistic steering behavior, while preserving minimal safety

distances and formations guidelines, according to tactical and

safety regulations, and staying in coherence with its local

formation and coordinated with whole hierarchy.

Index Terms—Multi-agent systems, hierarchical behavior,

crowd simulation, formation keeping.

I. INTRODUCTION

RMIES consist of huge numbers of units that are

formed into multi-level hierarchical structures that

follow the military rules for formations and movements. We

can consider the army as a large group comprised of sub-

groups, and every sub-group is a collection of other sub-

groups, until we reached the end of the hierarchy which is

the individual unit that can be a soldier, an armored vehicle,

a fighter, or a battleship. The brigade is made up of a group

of battalions, and the battalion is made up of a group of

companies, the company is made up of a group of platoons,

finally, the platoon is made of a number of individual units.

Commands are transferred to each group and sub-group

through the chain of commands sent from the group’s

Manuscript received July 10, 2012; revised July 12, 2012.

Abdulla M. Mamdouh is with Department of Computer Science, College

of Computing and Information Technology, Arab Academy for Science,

Technology & Maritime Transport (AASTMT), Misr El-Gadida, Cairo,

Egypt (Phone: +21227214675, +224934315; e-mail:

[email protected], [email protected]).

Ahmed Kaboudan is Lecturer Doctor at Department of Computer

Science, High Institute of Computer and Information Technology (HICIT),

Shorouk Academy, Cairo, Egypt, (Phone: +21002114631; e-

mail:[email protected], [email protected]).

Ibrahim F. Imam is Professor of computer science at Department of

Computer Science, College of Computing and Information Technology,

Arab Academy for Science, Technology & Maritime Transport

(AASTMT), Misr El-Gadida, Cairo, Egypt (Phone: +21222242929; e-mail:

[email protected] , [email protected]).

leaders. Those commands propagate through a network that

connects the leaders with their units. At every level of the

hierarchy, the group at this level tries to keep coherent to his

behavior.

It is a big challenge to model a 3D virtual environment

that simulate in real-time the movements of armies with the

previous mentioned complex structures and constraints, as

this type of models has a strong relation to many fields of

computer science like artificial intelligence, computer

graphics and robotics. In this paper we model a system that

simulates the movements of complete brigade of tanks in

different regular and combat situation on arbitrary terrain

topologies, while designing our algorithms and techniques

to cope with real-time applications requirements. We deal

with the following problems in this paper:

• Modeling multi-level hierarchical autonomous

steering behavior for group of tanks with target seeking,

avoiding obstacles and neighbors.

• Compounding formations distributed on the

groups and sub-groups to the leaves units while keeping

coherent with each other group during the movement. The

formations follow a set of arbitrary military rules.

The crowd simulation is a topic that was investigated

since the 1980s when the Reynolds introduce the boid

system to model the motion of flock of birds, herds and land

animals or a schools of fishes[3].Many problems are

encountered when simulating a group of moving entities.

For instance, predicting the future collision with obstacles

and trying to avoid them, avoiding collision with moving

neighbor units, also passing through narrow corridors is a

very important issue. These problems should be dealt with

while taking into consideration the alignment, velocity and

keeping the coherence of the group. We will be using the

vehicle as the individual unit.

The formation conservation is a very interesting topic for

crowd simulation and multi-robot system. It has wide usages

in the military applications as almost all movements are

restricted with special formations rules that are designed to

increase the probability of hitting the targets and enemies

and it also reduces the rate of destruction of the formatted

forces. In addition it is used to increase the range of vision

of the group in tactical movements. Many issues should be

considered with moving formation groups like the effect of

obstacles that distort the group formations. Dynamic

switching, which mean changing the formation styles during

the movements, also increases the complexity of the

Realistic Behavioral Model for Hierarchical

Coordinated Movement and Formation

Conservation for Real-Time Strategy and War

Games

Abdulla M. Mamdouh, Ahmed Kaboudan, and Ibrahim F. Imam

A

IAENG International Journal of Computer Science, 39:3, IJCS_39_3_03

(Advance online publication: 28 August 2012)

______________________________________________________________________________________

mailto:[email protected]



problem.

In our model we are mixing the crowd simulation with a

formation that is composed of sub-formations distributed

among a multi-level of hierarchy of brigade structure. The

brigade hierarchy, is composed of battalion, companies,

platoons sub-groups respectively and end with tanks, which

represent the virtual autonomous agents. These virtual

agents have different level of autonomy. They behave

according to the distributed orders and organized with

standard formation which is distributed among the multi-

level tree structure of the groups while trying to follow the

internal behaviors that are fed to the agents through their

own sensors.

In contrast to crowd simulation, swarms, flocks and other

well studied mass units’ movements, which mainly try to

follow a path or goal, avoid obstacles and keep a

predetermined distance between each other, the movement

of military units is much more complex and involves many

planning and coordination decisions.

The top level of the hierarchy, the brigade in our case,

propagates strict goals down the chain of command. While

these goals should be followed by all units, real gaming

situations mandate some temporary deviations from the

planned goals. Then it is up to each sublevel to decide how

to manage the current situation, and re-plan a temporary

strategy to overcome this situation. The temporary change of

orders should then again broadcast to sublevels, until we

reach the individual unit, which is also should handle and

plan its movement according to the current situation.

Finally, all levels should recuperate its primary goal and

restore the movement accordingly. In this paper we will

implement our model, presented in previous paper [23], and

present a fully functional program portraying all the

forgoing concepts.

The rest of this paper is organized as follow. Section II

gives an overview of the related work for group movement,

coordinated movement with formation structure and

keeping, vehicles motion simulation and military movement,

including our contribution. The next section III discusses

our model to simulate a group of tanks movement with all

applied constraints like formation and coherence,

hierarchical structure and collision avoidance. Also present

the architecture model for our system and the techniques

used to implement the hierarchical behaviors. In section IV

we present the group movement and the types of movement

of group of tanks to achieve its tactical mission, also present

our techniques and approaches to implement that movement.

Section V discusses our techniques to implement the multi-

level hierarchical formation. Finally, in section VI we

conclude the paper and talk about future work.

II. RELATED WORK AND CONTRIBUTION

Reynolds [3] introduced one of the most common

approaches to simulate the group movements. He proposed a

boid-model that simulates the non-colliding aggregate

motion which is created by distributing the three basic

behaviors (coherence, alignment, and separation) to his

flock, herds, and schools. Reynolds inspires his approach

form the particle system [6]. This model use local rules for

individual to make the flock be together and avoid collision

with each other and with other objects. In 1999 Reynolds [4]

extended his technique to simulate in real-time the steering

behavior of autonomous agent and combining the basic

steering behavior to achieve more complex behavior for

movements. Then he implemented the crowd simulation of

thousands of agents moving on the PS3®

platform using the

new parallel architecture where he used the spatial hashing

for space partitioning and for optimizing the performance

and accelerating the crowd simulation [5].

Another approach used for modeling crowd simulation is

based on continuum perspective rather than per agent-

dynamics. This approach proposed by Hughes [7] for

modeling crowd motion using a potential field function.

Treuille [8] developed the Hughes model to integrate the

global planning and local collision avoidance to simulate the

motion of large crowds without needing for explicit

collision avoidance.

In robotics, there are many multi-robot system researches

for modeling the motion of groups of robots. Li [9] proposed

a centralized path planner based on spherical tree hierarchy

structure that supports dynamically grouping of robots.

However, this approach just reduces the inter-robot collision

and does not guarantee the coherence of the group of all

robots.

Kamphuis [10] created a method to keep the coherence of

groups during movements. This approach is designed to

generate a single path as a backbone to a corridor using the

clearance algorithm along the path. This guarantees the

TABLE I

COMPARISON BETWEEN DIFFERENT TECHNIQUES [17] [23]

Formation Coherence Narrow

passages Performance

Hierarchical

behavior Individual type in crowd

Pottinger [12], [13] Always No Natural Excellent No Character

Balch [14] Always No Not informed Excellent No Virtual robot

(motor-based)

Li [9] Always No Not-natural Excellent No Virtual robot

(humanoid)

Kamphuis [10] No Always Not-natural Excellent No Character

Berg [21] No No Not-natural Good No Character

Silveira [17] Yes

(configurable)

Yes

(configurable) Natural Good No Character

Our Technique

Yes

(configurable)a

(multi-level)

Yes Natural Goodb

Yes

(multi-level) Vehicle

aConfigurable formation in our techniques means that the group can follow the new formation presented by its commander while moving and also the

formation can be changed at every time the group decides to move with or without any constraints. bParallel implementation is planned for future work.



______________________________________________________________________________________

coherence of the group as all the units can move inside the

corridor without splitting by the cluttered environment.

Although this technique keeps the coherence but lacks

formation, collision avoidance, and it is not applicable for

large number of units.

Almost all group motion animations are modeled for

organics like humans, animals, birds and fishes and for

multi-robot systems. Jared [11] present a method for

autonomous interactive vehicles moving in 2D and 3D

environments such as air, water, and land. This system

combines the steering behavior including seek, pursue,

avoid collision, and flee with real-time path planning.

Pottinger [12], [13] proposed a coordinated movement

model with predefined formation supporting collision

avoidance of a group of units to be used in real-time strategy

(RTS) games. However, his technique does not support

dynamics switching, or scalable formations. Balch [14]

designed a new approach to arrange multiple robots in

geometrical formations using a new class of potential

function while moving toward a goal position and avoiding

collision with obstacles. The potential function generates

many types of forces; one is used to force the robots to

move toward the attachment sites that are located around the

robot while other one is used as repulsion force from the

center of obstacle to make a robot go away from obstacle

during the movement toward the goal.

Musse and Thalmann [1] proposed the ViCrowd model to

simulate a hierarchical crowd structure and distributed the

behaviors through three levels of the crowd, composed of

groups and agents with different degree of autonomy.

Boccardo [2] developed a framework called Massive Battle

that simulate the coordinated movement of the soldiers

grouped in the platoons that are marching along a path and

engaging in a combat. The system also presents the courage

factor of the soldier during fighting.

Trevisan and Dapper [15], [16] using the boundary value

problem (BVP) which is a class of potential differential

equation (PDE). Trevisan used this technique to allow

robots to navigate and explore the unknown environment,

Dapper used the technique to steer synthetic actors to move

while avoiding the collisions and attaining goals. Silveira

[17] extended the Dapper (BVP) approach to manage the

movement of a group while keeping the formation with any

desirable shape and preserve the configurable coherence of

the group that can handle the collision avoidance with

obstacles. They implemented this approach to work on

multi-core CPU and GPU.

Mamdouh [23] proposed the new model to handle these

previous problems while applying the formation and the

coherence of the coordinated groups on multi-level

hierarchical structure following arbitrary organization of

armies. The system is modeled using a rule-based multi-

agent system to achieve the required intelligence of the final

movement of individual units. Table I was modified form

the original version [17] to show our techniques. It

compares between previously presented techniques with

ours considering the formation conservation, coherence

keeping, crossing from narrow areas, hierarchical behavior

distribution, the performance and the type of individual unit

used by group. Table II based on [25] and the modified

version [1] which presents various approaches used in the

crowd simulation modeling and compare between our

method with others based on the structure of the group,

number of individuals in the crowd, the level of intelligence

of the output behaviors, using of physics in the simulation,

types of collision avoidance, and the style of controlling the

group behaviors.

III. SIMULATION MODELING

As we show in Fig. 1, which presents the flow of data that

is used to simulate the group of tanks formed into

hierarchical structure of the military standard. In real-life as

well as in our system, the commander of operations inter the

global path on the passable area on the presented map while

he neglect the small detail like any obstacles that can be

TABLE II

VARIOUS APPROACHES USED TO MODEL CROWD BEHAVIORS [25] [1]

Method Particle

Systems

Flocking

Systems

Behavioral

Systems

Our

Methods

Structure Non-

hierarchical

Levels:

flock,

agents

Can

present

hierarchy

Multi-level

of

hierarchy

Participants Many Some Few Some

Intelligence None Some High High

Physics-

based Yes Some No Some

Collision Detect&

respond Avoidance Avoidance Avoidance

Control

Force

fields,

global

tendency

Local

tendency Rules

Predefined

behaviors

and Rules

Fig. 1 Flow of data of group movement with hierarchical structure[23].



______________________________________________________________________________________

trees, buildings, big stones and rocks. During the simulation,

the commander will create a global path by selecting a

sequence of multiple points on the grid-based map. The line

segments that connect these points will form the global

path. We will call these connection points as checkpoints.

These checkpoints are a 3D positions plotted on the terrain.

The group of units will consider the sequence of these

checkpoints as targets that will be reached sequentially one

after the other. When the group reaches a target checkpoint,

the next checkpoint will be the new target position that must

be followed by the group. And that is done until the final

checkpoint is reached.

There are many forces that applied on the group to

achieve the desired intelligent movement of group. Global

path following forces coordinate the group to move on the

path while the multi-level hierarchy formation forces

applied to make the group keeping its formation and be

coherent, for more details about the formation see section

[Formation]. Obstacle avoidance forces make the individual

tank in the group move away from the obstacles that were

not included during the global path creation. Obstacles are

detected by virtual vision sensors that are attached to each

tank agent. Tanks avoid the collision with each other by

using the neighbor collision avoidance forces generated by

using the virtual vision sensors. The virtual agents that

coordinate the movements of tanks use a pre-defined rule-

based approach to select or mix the appropriate forces

according to the current situation to pass forces to vehicle

steering system to generate the tanks motions.

Commander tank agents can send orders to individual

tanks through a virtualized network that is structured in a

hierarchical topology that follows the structure in Fig. 2.

This topology forces the flow of commands to be directed to

specific agent(s). Also it allow individual tank agent to

communicate with its leader without conflicting with other

leaders.

The group of tanks is coordinated by the global movement

strategy which controls the type of the current movement

methodology. This global movement strategy describes the

behavior of the group movement by changing the formation

and the style of movement which is achieved by converting

the global goals and internal targets for each agent. These

movement strategies are controlled by a finite state machine

[18]. See Fig. 4 to see the state machine that controls the

global movement strategy. Vehicle steering behavior system

receives the final force vector after applying the rules of

intelligent movement like following formation and avoiding

collision with obstacles and neighbors. This vector is used to

generate the final velocity of the tank using the steering

behavior introduced by Reynolds [4], [19], [20]. We choose

to use and adapt the steering behavior as it has a good

performance for computing physically-based model of a

simple moving vehicle. Finally this approach generates the

new orientation of the tank which is used to update the

Fig. 2 Hierarchy structure of brigade.

A Platoon is composed of three tanks, a Company is composed of three

platoons plus the company’s leader (10 Tanks), a Battalion is

composed of three companies plus the battalion’s leader (31 Tanks),

finally a Brigade is composed of three battalions plus the brigade’s

leader (94 Tanks).based on NATO symbols [22] [23].

Fig. 3 Architecture model: static structure for the system components.



______________________________________________________________________________________

position in real-time loop. At the last stage in our simulation

we adapt the orientation of the tanks on the terrain

topography and then link the result data with the 3D

graphics engine to visualize the result.

Fig. 3 show the static structure of the architecture model

of our system. It contains the major components that

compose the system. IStrategy component encapsulate the

three type of global movement strategies. Hierarchical thre

component contains the ternarty tree data structure that used

to model the hirarchical structure to use for comunication

between the agents and formation creation for multi-level

groups and sub-groups. The group controller is the virual

agent that control the sub-group behaviors. Formation

manager encapsulate the functions that handle the fomration

creaion and control the movement of target formation

points, see section [Formation] for more details. 3D

environemet component includes the global path that will be

followed by the whole group and the terrain manager that

used to handle the terrain information. Finally the virtual

agent is the compononet that describe the individuals in the

group which is composed of the tank agent with the steering

behavior and the vision sensors that used for avoiding the

collision with dynamic and static objects.

A. Hierarchical Structure

To map the hierarchical structure of the military to be as

shown in figure we create a tree data structure called ternary

tree. The ternary tree contains exactly three child nodes for

all nodes in the tree. This tree helps us to map the presented

military model which involves three subordinates for each

commander at each level. We use nodes to encapsulate the

adjacent individual tank in the brigade, the formation of the

subordinates and the target position that should be followed

by the tank to achieve its goal and following the formation,

see section [Formation]. The edges in the tree force the

group to be organized and subdivided into sub-groups to

follow the hierarchical structure while determining the

virtualized network topology to be used to carry the order

messages from leaders to the subordinates. This model

allows controlling the sub-groups behaviors directly by

sending a direct message to node which encapsulate the sub-

group leader and then broadcast the order message to

subordinates respectively.

B. Hierarchical Behavior

Our model supports multi-level of group controlling by

using a group controlling agent to control the whole group,

multi-level sub-groups, and the individual agent tanks.

Whole group is controlled by distributing the global

movement strategy to all tanks. The group controlling agent

can create the sub-group controller and assign the specific

behavior to it and determine the new goals and sub-group

leader. The sub-groups are controlled to change the

movement direction, speed, and status from moving to

stopping. The sub-group controller can divide again into

new sub-group controller in different level which supports

the multi-level of hierarchical behavior. Each sub-group

split to move as a separate group following another specific

goal while keeping the group coherent. That is done after

receiving an order from its commander.

Fig. 6 Opening and spreading movement sequence.

Fig. 5 Hierarchical behavior structure of the model.

Fig. 4 Global strategy state machine.



______________________________________________________________________________________

At the end, the individual tanks execute the global

movement strategy by using the steering behavior [4] with

individuals’ goals that form the global movement strategy.

Also individual tank use the attached rules to automatically

avoid the collision with obstacle and neighbor tanks while

keeping the formation and coherent. See Fig. 5 that show the

hierarchical behavior distribution for multi-level structure.

IV. GROUP MOVEMENT

We define the group based on Musse and Thalmann

definition [1] as a collection of similar entities that behave

coherently to achieve a common goal in the same virtualized

environment. We consider the army as a large group

composed of a hierarchical tree structure as shown in Fig. 2.

In our model we simulate the movements of group of tanks

collected as a full brigade. The movement of the tanks can

be split into three types. The first type of tank movements is

for transportation, used when a group of tanks needs to

move from current position to another target position

without any enemy engagement. In this case the group

moves in queue formation following his leader. The second

type of movements occurs when the group change its status

from queue formation to opening by spreading and each

tank tries to move toward specific direction to become in a

fighting formation and this is done down through each level

within the hierarchy in the group through many sup-phases.

The last movement type is the movement for engagement. In

this situation the group moves in a multi-level hierarchical

formation that allows individual tanks to better engage with

the enemy in different formations according to the combat

situation.

A. Transportation Movement

As we mentioned before, we need the transportation

movement to move a group from one place to another. That

happens without engagement at any combat. In this type of

global movement strategy the tanks are queued into line

formation to move on the global path. Each tank in the

platoon group considers the front unit as a target to follow,

while keeping a pre-configured distance between them. At

the next level of the hierarchy, the first tank in the company

group tries to keep the specific distance from the last front

tank in the front company. This is repeated in each level in

the battalion and brigade.

Each agent in the group uses the following rule to keep

the fixed distance between each other. This rule forces the

agent to slow down its speed when the distance between the

tank and front tank is less than the pre-defined specific

distance and increase the speed when the distance becomes

larger than the specific distance.

B. Opening and Spreading Movement

This type of movement represents the transition phase

between the transportation movements and the engagement

movements. In these strategic movements the hierarchical

sub-groups will split and change their directions through

many phases to achieve the final desired formation and be

ready to engage in combat according to specific situation.

Each stage bounded by two opening lines: lines

perpendicular to the global path. The first opening line

determines the position and the group level in the hierarchy

that should be starting to split, changing the direction of its

sub-groups. The second opening line contains the target

positions that should be followed by the divided groups.

These target positions follow the desired final formation of

the brigade. This formation leads to setting the target

positions to the final opening line. Group splitting follows

the following strategy: the brigade is divided into three

battalions while each battalion moves toward a specific

direction. The same operation occurs sequentially with each

sub-group starting form battalions, companies, platoons and

ending with individual tanks. To achieve this movement

behavior, each tank leader starts to change his direction by

following the new target on the second opening line in his

stage. The rest of the tanks belonging to its group follow

their leader in the queue formation with the same movement

behavior as described in the transportation movement

section. See Fig. 6 that show the steps of the opening and

spreading movement.

C. Engaging Movement

In this strategic movement, the group of tanks is moving

in a specific formation that is distributed through multi-level

hierarchy structure. Tanks try to follow the global path

while it keeps its formations. In all three types of

Fig. 7 Some platoon formations [23].

Fig. 8 Battalion formations [23].



______________________________________________________________________________________

movements the individual units rely on internal rules to

avoid collision with obstacles and neighbors while following

the global path and keep the coherence of the group.

V. FORMATIONS

As mentioned earlier, the formation of the units is crucial

for military movements. It is used for achieving tactical

goals during engagement by increasing the effectiveness of

fighting units. Our model supports the formation through the

hierarchical multi-level structure. This means that formation

is composed of sub-formations, and the sub-formations are

composed of other sub-formation as shown in Fig. 9.There

are many types of formation at every level in the hierarchy

of military structure. The platoon formation can be arranged

in six types of formations, as shown in Fig. 7. At the

battalion level there are four types as we can see in Fig. 8.

Fig. 9, shows a battalion formation composed of two

arrangement types while each company in the battalion is

composed of three platoons. The top left company in Fig. 9,

shows the three platoons formed from left to right into right

flank, wedge, and vee formations.

Our approach to keep the formation during movement is

inspired from the carrot and stick principle. In our model

there are target formation points organized into the form of

desired formations of the whole group and sub-groups. All

of this target formation points will move along the global

path and the tanks will try to follow this points. Every tank

will be attached to specific target point within its group

formation.

This technique supports the dynamic switching between

any types of formation without the need of extra logic, e.g.

as when we change the formation of any group which leads

to changing the corresponding target formation points which

in turn leads the tanks to follow the new position of target

formation points.

We solved the problem of crossing through narrow area

by adding two wing sensors for detecting the obstacles at the

left and right edges of the platoon. When both sensors detect

forward obstacles at the same time, the platoon will change

its formation to line formation and use the same technique in

the transportation movement to be queued while moving, as

shown in Fig. 10. After the platoon passes from the narrow

area it returns back to its original formation again. This

technique ensures the passing between any obstacles and

avoids neighbors from collision while the motion is very

natural.

A. Formation Creation

To create the desired formation the user must provide the

target formation for each group at each level through the

hierarchy. Then each tank at each group will be assigned to

only one target formation point. We use a ternary tree data

structure to set the target formation point positions. That is

done by traversing the tree breadth first and at each node the

algorithm reads the current node formation and assigns the

relative positions for target formation points to the children

nodes.

B. Formation and Coherence keeping

Every tank has a three speed state (normal speed,

maximum speed, and minimum speed).We use a rule-based

behavior that is built in our code to maintain the cohesion

and formation of the group during movements, and forcing

each tank to follow its target formation point. Then using

the following rules is that if the tank becomes too close or

too far from its target formation point, due to imposed

situation, the rules tries to decrease or increase the speed,

using variable threshold acceleration, but always respecting

the minimum and maximum defined speeds of the specific

tank type. Show the Fig. 11 to see the formation and

coherence keeping while moving and avoiding the collision.

VI. CONCLUSION AND FUTURE WORK

We presented a model to simulate in real-time the

movements of group of military vehicles (tanks) using rule-

based multi-agent system, which is suited to model the

behavior of the group and make the movements more

natural in 3D virtual environment. Our model is mixing the

global path following with local steering behavior to

achieve final group movement. Another contribution of our

work is the modeling of multi-level hierarchical structure

and the distribution of intelligent movement behavior and

formation conservation through this structure while keeping

the coherence of groups. Our model supports dynamic

switching between formations according to strategic

situations, while avoiding collision with neighbor units and

obstacles. We investigate the simulation of group movement

through three types of global movement strategies:

Fig. 10 Platoon passing through narrow area [23].

Fig. 9 Battalion multi-level hierarchical formations [23].



______________________________________________________________________________________

transporting, spreading, and engaging movements of the

group.

We use the Unity3D [24] as a graphics engine to

implement a visual part of our system. Also we used the

OpenSteerDotNet library [20] which is the port version of

the OpenSteer library [19] to generate the basic steering

behaviors. We test our model on an off-the-shelf PC: intel®

i5-2500 3.2 GHz, 8 GB RAM, GeForce 465 GTX with 1GB

RAM. The system can handle the behaviors of the full

brigade - as we described before - with around 100 units of

tanks on an interactive frame rate (230) FPS (frames per

second).

For future work, we plan to enhance the performance

through parallel implementation of our algorithms.

ACKNOWLEDGMENT

The authors are grateful to Mohammed Mamdouh and

Eid Samy for valuable help and to the anonymous reviewers

for constructive criticism.

REFERENCES

[1] S. R. Musse, and D. Thalmann, “Hierarchical model for real time

simulation of virtual human crowds”, IEEE Trans. Visual Comput.

Graphics, vol. 7, no. 2, pp. 152 – 164, Apr. 2001.

[2] A. Boccardo, R. D. Chiara, and V. Scarano, “Massive battle:

coordinated movement of autonomous agents”, CASA Workshop on

3D Advanced Media in Gaming and Simulation, Amsterdam and

Netherlands, Jun. 2009, pp. 40–47.

[3] C. W. Reynolds, “Flocks, herds and schools - a distributed behavioral

model”, Proc. 14th annu. Conf. on Computer graphics and interactive

techniques, New York, USA, Jul. 1987, pp. 25–34.

[4] C. W. Reynolds, “Steering behaviors for autonomous characters”,

Proc. of Game Developers Conference, California, USA, 1999, pp.

763–782.

[5] C. W. Reynolds, “Big fast crowds on PS3”, Proc. 2006 ACM

SIGGRAPH symposium on Videogames, New York, USA, pp. 113–

121.

[6] W. T. Reeves, “Particle Systems-A Technique for Modeling a Class

of Fuzzy Objects”, ACM Trans. on Graph. (TOG) Proc. ACM

SIGGRAPH 1983, vol. 2, no. 2, pp. 359-376.

[7] R. L. Hughes, “The flow of human crowds”, Annu. Rev. Fluid Mech.,

vol. 35, pp. 169–182, Jan. 2003.

[8] A. Treuille, S. Cooper, and Z. Popovi´c, “Continuum crowds”, ACM

Trans. on Graph. (TOG) Proc. ACM SIGGRAPH 2006, vol. 25, no. 3,

pp. 1160–1168.

[9] T. Li, and H. Chou, “Motion planning for a crowd of robots”, IEEE

Trans. Robot. Autom. , vol. 3, pp. 4215–4221, Nov. 2003.

[10] A. Kamphuis, and M. H. Overmars, “Finding paths for coherent

groups using clearance”, Proc. 2004 ACM SIGGRAPH/Eurographics

Symp. on Comput. Animation, Aire-la-Ville, Switzerland, pp. 19–28.

[11] J. Go, T. Vu, and J. J. Kuffner, “Autonomous behaviors for

interactive vehicle animations”, Proc. 2004 ACM

SIGGRAPH/Eurographics Symp .on Comput. Animation, Aire-la-

Ville, Switzerland, pp. 9–18.

[12] D. Pottinger, (1999, Jan.), Coordinated unit movement, Game

Developer Magazine, pp. 42–51.

[13] D. Pottinger, (1999, Feb.), Implementing Coordinated Movement,

Game Developer Magazine, pp. 48–58.

[14] T. Balch, and M. Hybinette, “social potentials for scalable multi-robot

formations”, IEEE Int. Conf. Robotics and Automation, 2000 Proc.

ICRA '00. , vol.1, pp. 73 – 80.

[15] M. Trevisan, M. A. P. Idiart, E. Prestes, and P. M. Engel,

“Exploratory navigation based on dynamical boundary value

problems”, J. Intell. Robot. Syst., vol.45, no. 2, Feb. 2006, pp. 101–

114.

[16] F. Dapper, E. Prestes, and L. P. Nedel, “Generating steering behaviors

for virtual humanoids using BVP control”, Proc. Comput. Graph. Int.

(CGI), 2007, pp. 105-114.

[17] R. Silveira, E. Prestes, and L. P. Nedel, “Managing coherent groups”,

Comp. Anim. Virtual Worlds, vol. 19, no. 3-4, Aug. 2008, pp. 295-

305.

[18] I. Millington, and J. Funge, “State machines”, in Artificial

intelligence for games, 2nd ed. M, Burlington, Morgan Kaufmann,

2009, ch. 5, sec. 3, pp. 309–331.

[19] OpenSteer, [Online]. Available: http://opensteer.sourceforge.net/

[Accessed: Oct. 25, 2011].

[20] OpenSteerDotNet. [Online]. Available:

http://code.google.com/p/opensteerdotnet/ [Accessed: Oct. 25, 2011].

[21] J. Berg, S. Patil, J. Sewall, D. Manocha, and M. Lin, “Interactive

navigation of multiple agents in crowded environments”, SI3D’08

Proc. Symp. Interactive 3D Graph. and Games ACM, New York,

USA,2008, pp. 139-147.

[22] D.U. Thibault. (2005, Sep.). Commented APP-6A - Military symbols

for land based systems NATO’s current military symbology standard.

Defence R&D Canada – Valcartier. Canada. [Online]. Available:

http://www.mapsymbs.com/APP-

6ADRDCValcartierEdition121(Mod).pdf

[23] A.M.Mamdouh, A.Kaboudan, and I.F.Imam. “Real-time, Multi-agent

Simulation of Coordinated Hierarchical Movements for Military

Vehicles with Formation Conservation”, Proc. of Int. Multi Conf. of

Eng. and Comput. Scientists, IMECS 2012, Mar. 2012, Hong Kong,

pp. 139-147.

[24] Unity3D. [Online]. Available: http://www.unity3d.com/ [Accessed:

Jul. 11, 2012].

[25] R.Parent, “Advanced Algorithms”, in Computer Animation

Algorithms and Techniques, San Francisco, USA, Morgan Kaufmann,

2002, ch. 4, sec. 5, pp. 241–260.

Fig. 11 Formation keeping (a) Company in the final three arrangement formations and the formations of platoons from left to right are (left flank, line,

and left flank); (b) The tanks follow the global path and avoiding the collision with obstacles while trying to keep the formations and be coherent;

(c) Tanks following its target formation points.



______________________________________________________________________________________

http://opensteer.sourceforge.net/

http://code.google.com/p/opensteerdotnet/

http://www.mapsymbs.com/APP-6ADRDCValcartierEdition121(Mod).pdf

http://www.mapsymbs.com/APP-6ADRDCValcartierEdition121(Mod).pdf

realistic behavioral model for hierarchical coordinated ... · abstract—in real-time strategy...

Documents